Calculating Grid Partitioning Costs Of Distributed Virtual World Simulation Systems
Keywords
Distributed Simulation; Open Simulator; Partition Model; Virtual Worlds
Abstract
Designing high fidelity and scalable distributed virtual environments remains a major challenge. Concurrent users, scene complexity and fidelity of user interaction are affected when computational resources become saturated because of high network traffic and heavy CPU load. Previous work dedicated to dynamic load balancing attempt to solve resource allocation based on avatar count and activity by developing load balancing policies that distribute server load by reas-signing avatars to evenly distribute computational load. The efficient assignment of the system resources based on workload is considered a partitioning problem. Few studies have focused on developing cost models for the partitioning problem beyond avatar position and movement or from real world performance. In our research we have found that there is a difference in performance and CPU load between simulated agents (bots) and actual user controlled agents (avatars) not specifically related to agent count. In this paper we focus on presenting a cost model based on simulator specific factors that impact simulation performance using the OpenSimulator framework.
Publication Date
1-1-2015
Publication Title
Simulation Series
Volume
47
Issue
2
Number of Pages
25-32
Document Type
Article; Proceedings Paper
Personal Identifier
scopus
Copyright Status
Unknown
Socpus ID
84937777962 (Scopus)
Source API URL
https://api.elsevier.com/content/abstract/scopus_id/84937777962
STARS Citation
Rivera, William A.; Goel, Amit; Kincaid, Peter J.; and Karwowski, Waldemar, "Calculating Grid Partitioning Costs Of Distributed Virtual World Simulation Systems" (2015). Scopus Export 2015-2019. 1909.
https://stars.library.ucf.edu/scopus2015/1909